Re: pg_upgrade bug found!

Jeff Davis wrote:
> On Thu, 2011-04-07 at 15:46 -0400, Bruce Momjian wrote:
> > OK, so the only other idea I have is to write some pretty complicated
> > query function that does a sequential scan of each toast table and pulls
> > the earliest xmin/xmax from the tables and use that to set the
> > relfrozenxid (pretty complicated because it has to deal with the freeze
> > horizon and wraparound).
>
> That sounds like the correct way to fix the situation, although it's a
> little more work to install another function just for this one-time
> purpose. TransactionIdPrecedes() should already account for wraparound,
> so I don't think that it will be too complicated (make sure to read
> every tuple though, not just the ones currently visible).

I want to avoid anything that requires a compile because they are hard
for many sites to install so TransactionIdPrecedes() is out. We will
need to do this in PL/pgSQL probably.

> Stepping back a second to make sure I understand the problem: the only
> problem is that relfrozenxid on the toast table after an upgrade is
> wrong. Correct?

Yes, it was not restored from the old cluster.

--
Bruce Momjian <bruce(at)momjian(dot)us> http://momjian.us
EnterpriseDB http://enterprisedb.com

+ It's impossible for everything to be true. +